Agenda and Minutes, 15 November 2006
Agenda: Wed, 15 November 2006
Minutes
Attendance: Warren Anderson Peter Shawhan Stuart Anderson Ed Maros David Meyers Michael Samidi Greg Mendell Igor Yakushin Ben Johnson Brian Moe Kevin Flasch Adam Miller Nick Fotopoulos Patrick Brady (minutes) Kent Blackburn Announcements: Numpy and matplotlib is looking ok. Need to update LDG to 4.3 here at UWM. Worked fine here. FrGetVect issue that Drew reported was due to picking up the wrong python in an old version of LDG. Adam reported a problem with updating to 4.3 but got a success 4.0. Junwei needs to update documentation. Stuart reported fake entries for LDG users in /etc/shadow because of ssh strict check now. Patrick reported briefly on the CA and VOMS/GUMS discussions with Ruth and Kent last week. Actions to Ruth. DMT: No report. Glue: Duncan made modifications to the build instructions and user-env script; Adam has built an rpm. Patrick will test on jester. Nick suggested that we should ask people to send tests to Adam. Kipp mentioned that there are test suite. Igor has prepared a document to handle the changes to the segment database. He will put it into CVS and circulate a pointer. He mentioned that replication had problems because LIGOTools no report LAL/LALApps: nothing to report. LDAS: Testing with Opteron. Still some places cannot compile the code with Sun compiler. LDR: Nothing major architecturally to report. Since NEMO has come up, we have been asked to retrieve much data. Prioritized by the burst group. We have recently been requested to move the SFTs here. The ones the pulsar group wants here is to test E@H. Xavi is just transfering it now. Patrick to send e-mail to analysis committee about prioritizing data transfer. The question of data re-generation and re-transfering data is going to be an issue. Regeneration of h(t) real time should be able to keep up. How many MB of h(t) per clock second? How many MB per clock second from site to site? Stuart Kevin Murali Stefan Gerald Three pieces of technology will influence: small file issue; Greg ran some createRDS performance test show that switching compression mode; in last rounds of data server boxes that will be factors of several faster. MatApps: No report. People are having trouble connecting. Onasys: Significant maintenance work at the site yesterday. They have not yet started Onasys. Shourov reported that his cron jobs are still running. Binary inspirals, hoft and pulsar not running. Greg will enlist some help if it's not something simple. Stochastic online is not reporting any new segments, but he does see the scheduler universe. Greg will send e-mail about this if it is not an easy fix. OSG: David reported. In the last week have started new activities related to optimization. There are substantial data tranfer times needed. This is a major constraint. At a couple of sites in the future, we may be able to use SRM. In the short term, we may be able to set up level 4 caches. Will test out in a couple of weeks. Michael is trying to get the clustering done in the work flow. This is not working yet. Kent asked if this is the recommendation of the Pegasus team. Kent said that long term proposal is to move VDS out of the software stack. Then we would be responsible for moving it to $app. Have added two more sites for running hipe. Kent asked if they could pre-stage data anywhere yet. Level 4 has been staged at San Diego. They are using the UWM cluster. They are using many slots on the UWM cluster; Stuart pointed out that this raises the question of why we are doing this stuff. Michael reported 15 jobs running at most non-Milwauke sites. Have had ~100 at Milwaukee. Brian will get reporting on Ganglia for medusa. AOB:
$Id: 061115.html,v 1.2 2006/12/02 22:23:11 patrick Exp $